Recurring platform failure or unclear priorities
Reliability review
Failure map, ranked decisions and a practical sequence of work
Current focus
DevOps consultancy
•Production AI agents
•IoT / edge DevOps
Independent technical practice — Türkiye / worldwide
Platforms that deliver.Agents that know their limits.
I diagnose and improve cloud platforms, Kubernetes and delivery systems. I also design AI-agent workflows with evaluation, permissions, human review and operating costs built in from the start.
Reliable automation is designed around failure, evidence and human authority.
01 / Current practice
Work on the operating path,
not the slide deck.
Each engagement starts with one expensive uncertainty: a platform that keeps interrupting delivery, an incident that will repeat, or an agent workflow that cannot yet be trusted.
Primary practice01
DevOps & platform consultancy
Kubernetes, cloud infrastructure, Terraform, CI/CD, observability, DevSecOps and the developer path between them.
- Reliability review
- Kubernetes diagnosis
- Platform repair or build
Primary practice02
Production AI agents
One bounded workflow, a real baseline and the controls required to operate it without pretending the model is deterministic.
- Workflow pilot
- Evaluation and failure analysis
- Permissions, escalation and cost
Specialist track / 03
IoT & edge DevOps
Secure releases, containerised edge workloads, fleet observability and edge-to-cloud delivery. The work is operations-focused; I am not offering embedded product development.
Discuss an edge platform ↗03 / Evidence
One reported Kubernetes engagement
3 days→~2 hours
A customer described spending three days on the issue before the engagement. Diagnosis and resolution took roughly two hours. The public review record currently shows 20 five-star ratings from international customers.
Review the evidence and caveats ↗04 / Engagement shape
Small enough to decide.
Complete enough to keep.
Fixed scope where the problem allows it. Every proposal states the output, exclusions and how the result will be judged.
WhenWorkYou keep
One live or reproducible Kubernetes problem
Recovery session
Diagnosis, remediation, reasoning and prevention notes
One valuable manual workflow with a measurable baseline
Agent pilot
Working pilot, evaluation set, controls and rollout decision
An agent that works sometimes but is hard to govern
Agent review
Trace analysis, failure taxonomy, cost and control changes
05 / Perspective
Current work first.
History when it helps.
My practice today is DevOps consultancy and production AI-agent systems. Earlier work across Linux, electronics and IoT is useful where cloud operations meet connected fleets, but it is context—not the service proposition.
- Now
- DevOps, platforms, DevSecOps, agents
- Specialist
- IoT and edge operations
- Public record
- Writing since 2011, three technical books
- Teaching
- DevOps bootcamp, talks and workshops
06 / Writing
Technical work,
kept in public.
The archive is historical by design. Old posts remain at their original URLs, while current consulting and agent work stays at the front of the site.
Evaluating cloudnative-pg
Sep 23, 2023 / EN
↗Docker ile Redmine Kurulumu
10 Oca 2018 / TR
↗Statik bir Web Sitesini AWS CloudFront ve S3 kullanarak CDN Üzerinden Yayına almak
05 Oca 2018 / TR
↗Docker Multistage Build Örneği: Caddy Web Sunucusunun Dockerlaştırılması
28 Haz 2017 / TR
↗WordPress + MySQL with Isolated Network Deployment in 30 Seconds with Docker Swarm Mode
Jun 15, 2017 / EN
↗07 / Contact
What is failing,
slow or uncertain?
Send the current system, what has already been tried and the decision you need to make. Leave out credentials, customer data and confidential source code.